Testing merging two short 10 second clips made by an Xacti 1010, in 30 fps, 720P mode. Merged these two clips using MPEG Streamclip running on a Mac. This is testing if the derive using ffmpeg will fail on the audio as described in:
Audio/Visual sound, color
If it does fail, this would be evidence that it's the merging of two video clips using MPEG Streamclip that is causing the problem (since the problem does not seem to occur with videos shorter than 1 hour which do not have to be merged).